Uber India now offering gigs collecting info for AI models • The Register

Uber’s Indian arm has began utilizing its app to supply rideshare and supply drivers the possibility to make a Rupee by classifying knowledge utilized by AI techniques.

Megha Yethadka, international head of Uber AI Options, revealed the brand new gigs in a Thursday LinkedIn put up wherein she mentioned drivers generally have downtime throughout the day or may need to make some additional money after hours.

Yethadka mentioned the work can contain reviewing images, counting objects, classifying textual content, recording audio, or digitizing receipts.

She mentioned the gigs are “Powering our enterprise prospects worldwide for his or her gen AI fashions or shopper functions.”

“Till now, these duties had been accomplished by impartial contractors exterior the app,” Yethadka wrote. “The early outcomes are very promising, and we’re wanting to scale this additional.” In an accompanying video, she talked about “worldwide” growth for the providing.

Prabhjeet Singh, Uber’s president for India and South Asia, mentioned the gigs can be found in 12 cities and that “tens of 1000’s of drivers” are already performing what Uber calls “digital duties.”

The rideshare large’s CEO Dara Khosrowshahi talked about digital duties on the corporate’s early August Q2 earnings name, when he mentioned it is sensible for the corporate as a result of “It is utilizing the core Uber functionality, which is sending out duties to earners everywhere in the world. You’re simply going to see a unique sort of earner that’s going to work for the actually thrilling AI developments that you just see everywhere in the world.”

The posts don’t element how a lot Uber pays to finish digital duties or what it expenses prospects. We’ve requested the corporate for these numbers and can replace this story if we obtain a substantive reply.

Like Uber, however for knowledge lakes

Additionally on Thursday, Uber revealed it operates a 350-petabyte knowledge lake, and has created a device known as “HiveSync” to guard the info it accommodates.

“Uber’s batch knowledge infrastructure traditionally ran throughout two knowledge middle areas (main and secondary) to make sure redundancy,” explains a put up from the Uber Engineering staff. “Nevertheless, the secondary area sat idle—incurring prices equal to the first—simply to keep up excessive availability.”

Uber due to this fact launched a “Single Area Compute” (SRC ) program that sees all batch compute jobs run in a single area, earlier than HiveSync replicates the info to a second area.

The corporate constructed HiveSync by adapting an open supply undertaking known as ReAir created by Airbnb to duplicate tables and partitions between knowledge warehouses constructed on Apache Hive.

Uber began work on HiveSync in 2016. It now manages roughly 300 petabytes of information saved in 800,000 Hive tables and replicates eight petabytes of information every day.

“We plan to open-source this replication service and can proceed to develop new options to satisfy the rising calls for for scalability and decrease latency,” Uber’s put up states, including that HiveSync additionally performs “a essential position” as the corporate migrates batch knowledge analytics and ML coaching techniques to Google Cloud. ®